Continuous Delivery with Docker and Jenkins

Draining nodes

When we need to stop a worker node for maintenance, or we would just like to remove it from the cluster, we can use the Swarm draining node feature. Draining the node means asking the manager to move all tasks out of a given node and exclude it from receiving new tasks. As a result, all replicas are running only on the active nodes and the drained nodes are idle.

Let's look how this works in practice. Suppose we have three cluster nodes and a Tomcat service with five replicas:

$ docker node lsID                          HOSTNAME        STATUS  AVAILABILITY  MANAGER STATUS4mrrmibdrpa3yethhmy13mwzq   ubuntu-worker2  Ready   Active   kzgm7erw73tu2rjjninxdb4wp * ubuntu-manager  Ready   Active        Leaderyllusy42jp08w8fmze43rmqqs   ubuntu-worker1  Ready   Active   $ docker service create ...